Google Earth’s AI has just been withdrawn because it is flooded with deepfakes

Google deployed a new feature on July 30, allowing the Nana Banana 2 model to access the Google Earth map platform and edit location images with text descriptions. However, it was almost immediately misused.

Henk van Ess, an AI and open data researcher on the Internet, has created fake images with extremely high realism, such as “refugee concentration areas” along the US-Mexico border, or a complete nuclear power plant at an empty site in Iran.

Some social network users also use AI in Google Earth to create images of bogus floods or Godzilla monsters attacking downtown Los Angeles.

The above situation is difficult to control, forcing Google to withdraw the feature less than a day after deployment. “We’ve seen experts use the feature for useful purposes, but also found many people sharing fake photos and violating our policies. We decided to remove the feature from Google Earth while we develop stronger precautions,” a Google spokesperson said.

Google added that AI-generated photos are watermarked and are not shown on the Google Earth platform to ordinary users.